Martin recently refered me to a fascinating article and study, Self-Control at Work. Over 13 months, pay-by-the-field data-entry workers were offered the choice of a dominated contract, specifying that they would be paid half their normal rate if they failed to meet a daily target. These contracts would never pay more than a standard contract. Mathematically, they were strictly worse. Both imposed and self-chosen targets were experimented with. The authors claim “being offered the option to choose dominated contracts increases earnings by 9%.”¹

I’ve been familiar with this type of contract in the personal development realm for a while (say, for quitting smoking). They are typically called commitment contracts, and the change to dominated here is revealing.

The workers under study were already in a severely imbalanced power relationship with their employer.

Workers were told they were being hired for a one-time contract and were not provided reference letters upon completion of the job. […] Many employees commuted from surrounding villages using buses and trains, with some traveling up to two hours in each direction. […] Throughout the experiment, workers were randomly assigned to seats in the office and these assignments changed every one to three weeks.²

Dominated contracts further reinforce this power structure. They create a moral hazard for the company to sabotage work rates: if workers fail to meet targets they only need to be paid half as much. There is no evidence that this occured during the study, though I find it hard to believe this prospect would not eventually tempt the greedy hands of industry, if even unconciously.

A potential vector for malice is described in the paper. The assigned workstations were highly inconsistent, so much that the workers were able to identify the bad ones.

To test whether network uncertainty deterred workers from choosing targets the evening before work, we asked the office management staff to consult workers to identify which computers were more sensitive to network slowdowns. Management did not know the list would be used for this purpose. The computers identified as more uncertain are indeed more sensitive to overall network fluctuations

How easy it would be for a line manager to switch employee workstations at afternoon break, causing a dominated employee to miss their quota without substantially impacting output. (Even missing by one field results in penalty rates.)

That said, dominated contracts were emperically beneficial to many employees. Certain groups³ netted more income from the contracts, and generally expressed a desire for company assistance to maximize their economic well-being.

Some workers in our experiment asked us to withhold their earnings to help them save. Similarly, the nine members of the experiment’s managerial staff—who were paid fixed salaries—chose to receive their earnings monthly rather than weekly.

From within the existing power structure, this is certainly empowering. Extra income can lead to more opportunities (such as education, or capital for a business) and less dependence on employers.

Fortunately, a slight tweak to the contracts can retain the benefits while reducing the motivation for exploitation. If a contract is violated, the half-wage penalty should go to a third-party of the employee’s choosing (such as a charity), rather than the company. This is not a new idea: it’s a common arrangement used in research and has even been made into a website.

Employees who respond well to negative motivation could even choose an entity they dislike, though nominating the company itself should be banned to avoid obvious coercion opportunities. Since the dominated contracts never pay more than standard ones, there is no opportunity for the employee to collude with the third-party to increase their earnings. Both employer and employee can benefit from this arrangement without further reinforcing the employer’s power.

I’d also like to see deeper analysis into the effect of missing targets. What was the impact on output, motivation, retention, and happiness for workers that had to pay penalty rates? The study shows a 9% probability of missing a self-imposed target, with a large 12% standard error. How much of this was within the employee’s control? This could technically be measured quantitatively, but in general I feel qualitative interviews with workers would be an interesting complement to this study.

The author concludes mid-way through the article “we have not yet begun to scratch the surface of motivating production in this way”. This strikes me as upside down, and frankly kind of creepy. Motivating production!? Let’s talk of empowerment, agency, and people instead.